JP4977471B2 - Encoding apparatus and encoding method - Google Patents

Description

The present invention, audio signal relates to a coding apparatus and coding how that turn into codes an audio signal and the like.

  In order to effectively use radio resources and the like in mobile communication systems, it is required to compress audio signals at a low bit rate. On the other hand, users are demanded to improve the quality of call voice and realize a call service with a high presence. For this realization, it is desirable not only to improve the quality of the audio signal, but also to encode a signal other than audio such as an audio signal having a wider bandwidth with high quality.

  In response to such conflicting demands, an approach that hierarchically integrates a plurality of encoding techniques is considered promising. Specifically, a first layer unit that encodes an input signal at a low bit rate with a model suitable for a speech signal, and a residual signal between the input signal and the first layer decoded signal is also suitable for a signal other than speech. A configuration is adopted in which the second layer parts encoded by the model are hierarchically combined. An encoding method having such a hierarchical structure is called scalable encoding because the bit stream obtained from the encoding unit has scalability (a decoded signal can be obtained even from partial information of the bit stream). Because of its nature, scalable coding can flexibly support communication between networks with different bit rates. This feature can be said to be suitable for a future network environment in which various networks are integrated by the IP protocol.

As a conventional scalable coding, there exists a thing of a nonpatent literature 1, for example. This document describes a method for configuring scalable coding using a technique standardized by MPEG-4 (Moving Picture Experts Group phase-4). Specifically, in the first layer unit (base layer unit), a speech signal, that is, an original signal is encoded using CELP (Code Excited Linear Prediction), and in the second layer unit (enhancement layer unit) For example, the residual signal is encoded using transform coding such as AAC (Advanced Audio Coder) or TwinVQ (Transform Domain Weighted Interleave Vector Quantization). Here, the residual signal is a signal obtained by subtracting, from the original signal, a decoded version of the encoded code obtained in the first layer unit (first layer decoded signal).
Edited by Junichi Miki, “All about MPEG-4”, first edition, Industrial Research Co., Ltd., September 30, 1998, p. 126-127

  However, in the above prior art, the transform coding in the second layer unit is performed on the residual signal obtained by subtracting the first layer decoded signal from the original signal. Therefore, some of the main information included in the original signal may be removed by passing through the first layer unit. In this case, the characteristic of the residual signal is a characteristic close to a noise sequence. Therefore, for example, when transform coding designed to efficiently encode a musical tone signal, such as AAC or TwinVQ, is used for the second layer portion, the residual signal having the above characteristics is encoded to increase the decoding signal. In order to improve quality, it is necessary to allocate many bits. As a result, there is a problem that the bit rate is increased.

The object of the present invention has been made in view of this point, and a high-quality decoded signal can be obtained even if low bit rate encoding is performed in the second layer unit or higher layer unit. to provide a coding apparatus and coding how.

An encoding apparatus according to the present invention is an encoding apparatus that generates low frequency band encoded information and high frequency band encoded information from an original signal, wherein the low frequency band encoded information is reduced from the decoded signal. A first spectrum calculating means for calculating a first spectrum of a frequency band; a second spectrum calculating means for calculating a second spectrum from the original signal; and a filter having the first spectrum as an internal state. A first parameter calculating means for outputting a parameter indicating a characteristic as a first parameter indicating the degree of similarity between the first spectrum and the high frequency band portion of the second spectrum; and a spectrum in which a plurality of spectral residual candidates are recorded. the sign of the candidate of one spectral residuals from the residual shape codebook, variations formed between the first spectrum and the second spectrum of the high frequency band portion A second parameter calculating means for outputting as a second parameter indicating, among the first parameter and the second parameter is the output, the first to produce an estimate of the most similar to the high frequency band portion of the second spectrum A configuration is adopted that includes determination means for simultaneously determining a parameter and the second parameter, and encoding means for encoding the determined first parameter and second parameter as encoding information of the high frequency band.

The encoding method of the present invention is an encoding method for generating low-frequency band encoded information and high-frequency band encoded information from an original signal, wherein the low-frequency band encoded information is decoded from the decoded signal. A first spectrum calculating step for calculating a first spectrum of a frequency band; a second spectrum calculating step for calculating a second spectrum from the original signal; and a filter having the first spectrum as an internal state. A first parameter calculating step for calculating a parameter indicating characteristics as a first parameter indicating the degree of similarity between the first spectrum and the high frequency band portion of the second spectrum, and a spectrum in which a plurality of spectral residual candidates are recorded the sign of the candidate of one spectral residuals from the residual shape codebook, the first spectrum and the second spectrum of a high frequency band Generating a second parameter calculating step of calculating a second parameter indicating a fluctuation component, from the first parameter and the second parameter the calculated, the estimated value most similar to the high frequency band portion of the second spectrum and A determination step for simultaneously determining the first parameter and the second parameter, and an encoding step for encoding the determined first parameter and the second parameter as encoding information of the high frequency band. I made it.

  According to the present invention, a high-quality decoded signal can be obtained even when low bit rate encoding is performed in the second layer unit or higher layer unit.

  The present invention relates to transform coding suitable for an upper layer of scalable coding, and more specifically to an efficient spectrum coding method in the transform coding.

  One of the main features is that filtering processing is performed using a filter having a spectrum (first layer decoded spectrum) obtained by frequency analysis of the first layer decoded signal as an internal state (filter state), and an output signal thereof Is the estimated value of the high frequency part of the original spectrum. Here, the original spectrum is a spectrum obtained by frequency analysis of the delay-adjusted original signal. Then, the filter information for generating the output signal most similar to the high frequency part of the original spectrum is encoded and transmitted to the decoding unit. Since only the filter information needs to be encoded, the bit rate can be reduced.

In an embodiment of the present invention, a filtering process is performed by giving a spectral residual to the aforementioned filter using a spectral residual shape codebook in which a plurality of spectral residual candidates are recorded. In another embodiment, the error component of the first layer decoded spectrum is encoded before storing the first layer decoded spectrum in the internal state of the filter to improve the quality of the first layer decoded spectrum. The high-frequency part of the original spectrum is estimated by filtering processing. In still another embodiment, when the error component of the first layer decoded spectrum is encoded, the encoding performance of the first layer decoded spectrum and the estimation of the high frequency spectrum using the first layer decoded spectrum are performed. The error component of the first layer decoded spectrum is encoded so that both performances are high.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In each embodiment, scalable coding having a hierarchical structure composed of a plurality of layers is performed. In each embodiment, as an example, (1) the hierarchical structure of scalable coding includes a first layer (basic layer or lower layer) and a second layer (an enhancement layer or upper layer) that is higher than the first layer. (2) In the second layer encoding, encoding (transform encoding) is performed in the frequency domain. (3) The transform method in the second layer encoding is MDCT (Modified Discrete Cosine Transform). (4) In the encoding of the second layer, when dividing the entire band into a plurality of subbands, the entire band is divided into equal intervals on the Bark scale and each subband is divided into each subband. (5) F2 is greater than or equal to F1 (F1 ≦ F2) for the sampling rate (F1) of the input signal of the first layer and the sampling rate (F2) of the input signal of the second layer that correspond to the critical band It is assumed that there is a relationship.

(Embodiment 1)
FIG. 1 is a block diagram illustrating a configuration of an encoding apparatus 100 that forms, for example, a speech encoding apparatus. The encoding apparatus 100 includes a downsampling unit 101, a first layer encoding unit 102, a first layer decoding unit 103, a multiplexing unit 104, a second layer encoding unit 105, and a delay unit 106.

  In FIG. 1, an audio signal or an audio signal (original signal) having a sampling rate of F2 is given to the downsampling unit 101, and sampling conversion processing is performed in the downsampling unit 101 to generate a signal having a sampling rate of F1, The signal is provided to first layer encoding section 102. First layer encoding section 102 outputs the encoded code obtained by encoding the signal with sampling rate F1 to first layer decoding section 103 and multiplexing section 104.

  First layer decoding section 103 generates a first layer decoded signal from the encoded code output from first layer encoding section 102 and outputs the first layer decoded signal to second layer encoding section 105.

  Delay section 106 gives a delay of a predetermined length to the original signal and outputs the delayed signal to second layer encoding section 105. This delay is for adjusting a time delay generated in the downsampling unit 101, the first layer encoding unit 102, and the first layer decoding unit 103.

  Second layer encoding section 105 encodes the original signal output from delay section 106 using the first layer decoded signal output from first layer decoding section 103. Then, the encoded code obtained by this encoding is output to multiplexing section 104.

  The multiplexing unit 104 multiplexes the encoded code output from the first layer encoding unit 102 and the encoded code output from the second layer encoding unit 105, and outputs it as a bit stream.

Next, second layer encoding section 105 will be described in more detail. The configuration of second layer encoding section 105 is shown in FIG. Second layer encoding section 105 includes frequency domain conversion section 201, extension band encoding section 202, frequency domain conversion section 203, and auditory masking calculation section 204.

  In FIG. 2, the frequency domain transform unit 201 performs frequency analysis on the first layer decoded signal output from the first layer decoding unit 103 by MDCT transform to calculate MDCT coefficients (first layer decoded spectrum). Then, the first layer decoded spectrum is output to extended band coding section 202.

  The frequency domain transform unit 203 performs frequency analysis on the original signal output from the delay unit 106 by MDCT transform to calculate MDCT coefficients (original spectrum). Then, the original spectrum is output to extended band coding section 202.

  The auditory masking calculation unit 204 calculates auditory masking for each band using the original signal output from the delay unit 106 and notifies the extended band encoding unit 202 of the auditory masking.

  Here, human auditory characteristics include an auditory masking characteristic that when a signal is heard, even if a sound having a frequency close to that of the signal enters the ear, the sound is difficult to hear. The auditory masking is used to realize efficient spectral coding. In this spectral coding, quantization distortion that is permissible for auditory perception is quantified using human auditory masking characteristics, and a coding method corresponding to the permissible quantization distortion is applied.

  As shown in FIG. 3, the extension band encoding unit 202 includes an amplitude adjustment unit 301, a filter state setting unit 302, a filtering unit 303, a lag setting unit 304, a spectrum residual shape codebook 305, a search unit 306, and a spectrum residual. A gain codebook 307, a multiplier 308, an extended spectrum decoding unit 309, and a scale factor encoding unit 310 are included.

  The amplitude adjustment unit 301 includes the first layer decoded spectrum {S1 (k); 0 ≦ k <Nn} from the frequency domain transform unit 201, and the original spectrum {S2 (k); 0 ≦ k <Nw from the frequency domain transform unit 203. } Is given. Here, the spectrum score of the first layer decoded spectrum is expressed as Nn, the spectrum score of the original spectrum is expressed as Nw, and there is a relationship of Nn <Nw.

The amplitude adjustment unit 301 is configured such that the ratio (dynamic range) between the maximum amplitude spectrum and the minimum amplitude spectrum of the first layer decoded spectrum {S1 (k); 0 ≦ k <Nn} is the high frequency portion {S2 (k) of the original spectrum. The amplitude is adjusted so as to approach the dynamic range of Nn ≦ k <Nw}. Specifically, as shown in the following equation (1), the power of the amplitude spectrum is taken.

  Here, sign () represents a function that returns a positive / negative sign, and γ represents a real number in the range of 0 ≦ γ ≦ 1. The amplitude adjustment unit 301 uses the γ (amplitude adjustment coefficient) when the dynamic range of the first layer decoded spectrum after amplitude adjustment is closest to the dynamic range of the high-frequency part {S2 (k); Nn ≦ k <Nw} of the original spectrum. ) Is selected from a plurality of candidates prepared in advance, and the encoded code is output to the multiplexing unit 104.

The filter state setting unit 302 sets the first layer decoded spectrum {S1 ′ (k); 0 ≦ k <Nn} after amplitude adjustment to the internal state of the pitch filter described later. Specifically, the first layer decoded spectrum {S1 ′ (k); 0 ≦ k <Nn} after amplitude adjustment is substituted into the generated spectrum buffer {S (k); 0 ≦ k <Nn}, and the filtering unit 303 Output to. Here, the generated spectrum buffer S (k) is an array variable defined in the range of 0 ≦ k <Nw. A candidate of an estimated value (hereinafter referred to as “estimated original spectrum”) of the (Nw−Nn) point original spectrum is generated by a filtering process described later.

  In accordance with an instruction from the search unit 306, the lag setting unit 304 sequentially outputs the lag T to the filtering unit 303 while gradually changing the lag T within a predetermined search range TMIN to TMAX.

  The spectral residual shape codebook 305 stores a plurality of spectral residual shape vector candidates. Further, in accordance with an instruction from the search unit 306, spectral residual shape vectors are sequentially output from all candidates or from candidates limited in advance.

  Similarly, the spectral residual gain codebook 307 stores a plurality of spectral residual gain candidates. Further, in accordance with an instruction from the search unit 306, the spectral residual gain is sequentially output from all candidates or from candidates limited in advance.

  The multiplier 308 multiplies the spectral residual shape vector output from the spectral residual shape codebook 305 by the spectral residual gain output from the spectral residual gain codebook 307 to obtain the spectral residual shape vector. Adjust the gain. The gain-adjusted spectral residual shape vector is output to the filtering unit 303.

The filtering unit 303 performs a filtering process using the internal state of the pitch filter set by the filter state setting unit 302, the lag T output from the lag setting unit 304, and the spectral residual shape vector after gain adjustment. The estimated original spectrum is calculated. Here, the transfer function of the pitch filter is expressed by the following equation (2). Further, this filtering process is expressed as the following equation (3).

  Here, C (i, k) represents the i-th spectral residual shape vector, and g (j) represents the j-th spectral residual shape gain. The generated spectrum buffer S (k) included in the range Nn ≦ k <Nw is output to the search unit 306 as an output signal of the filtering unit 303 (that is, an estimated original spectrum). FIG. 4 shows the correlation among the generated spectrum buffer, the first layer decoded spectrum after amplitude adjustment, and the output signal of the filtering unit 303.

  Search section 306 instructs lag setting section 304, spectral residual shape codebook 305, and spectral residual gain codebook 307 to output lag, spectral residual shape, and spectral residual gain, respectively.

The search unit 306 also performs distortion E between the high-frequency part {S2 (k); Nn ≦ k <Nw} of the original spectrum and the output signal {S (k); Nn ≦ k <Nw} of the filtering unit 303. Is calculated. Then, a combination of a lag, a spectrum residual shape vector, and a spectrum residual gain when the distortion becomes the smallest is determined by an analysis method (AbS; Analysis by Synthesis) by synthesis. At this time, the auditory masking output from the auditory masking calculation unit 204 is used to select the combination with the smallest audible distortion. Assuming that this distortion is E, the distortion E is expressed by Expression (4) using a weight function w (k) determined by auditory masking, for example. Here, the weighting function w (k) takes a small value at a frequency where auditory masking is large (distortion is difficult to hear) and takes a large value at a frequency where auditory masking is small (distortion is easy to hear).

  The lag encoding code, spectral residual shape vector encoding code, and spectral residual gain encoding code determined by search section 306 are output to multiplexing section 104 and extended spectrum decoding section 309.

  In the above-described coding code determination method using AbS, the lag, the spectral residual shape vector, and the spectral residual gain may be determined at the same time, or each parameter is sequentially set (for example, the lag is reduced in order to reduce the calculation amount). , In the order of spectral residual shape vector and spectral residual gain).

  The extended spectrum decoding unit 309 includes an amplitude adjustment coefficient encoding code output from the amplitude adjustment unit 301, a lag encoding code output from the search unit 306, a spectral residual shape vector encoding code, and a spectral residual. The encoded code of the gain is decoded, and an estimated value (estimated original spectrum) of the original spectrum is generated.

  Specifically, first, amplitude adjustment of the first layer decoded spectrum {S1 (k); 0 ≦ k <Nn} is performed using the decoded amplitude adjustment coefficient γ according to the above-described equation (1). Next, the amplitude-adjusted first layer decoded spectrum is used as an internal state of the filter, and filtering processing is performed according to the above-described equation (3) using the decoded lag, spectral residual shape vector, and spectral residual gain, respectively. To generate an estimated original spectrum {S (k); Nn ≦ k <Nw}. The generated estimated original spectrum is output to scale factor encoding section 310.

  The scale factor encoding unit 310 includes a high-frequency part {S2 (k); Nn ≦ k <Nw} of the original spectrum output from the frequency domain conversion unit 203 and an estimated original spectrum {output from the extended spectrum decoding unit 309 { Using S (k); Nn ≦ k <Nw}, the scale factor (scaling coefficient) of the estimated original spectrum most suitable for hearing is encoded using auditory masking, and the encoded code is multiplexed by the multiplexing unit 104. Output to.

  That is, the second layer encoded code includes an encoded code (amplitude adjustment coefficient) output from the amplitude adjustment unit 301, and an encoded code (lag, spectrum residual shape vector, spectrum residual gain) output from the search unit 306. ) And the encoding code (scale factor) output from the scale factor encoding unit 310.

In the present embodiment, the extended band coding unit 202 is applied to the bands Nn to Nw, and a set of coded codes (amplitude adjustment coefficient, lag, spectrum residual shape vector, spectrum residual gain, scale) However, a configuration may be adopted in which the bands Nn to Nw are divided into a plurality of bands and the extended band encoding unit 202 is applied to each band. In this case, an encoding code (amplitude adjustment coefficient, lag, spectral residual shape vector, spectral residual gain, scale factor) is determined for each band and is output to the multiplexing unit 104. For example, when the bands Nn to Nw are divided into M bands and the extended band encoding unit 202 is applied to each band, M sets of encoded codes (amplitude adjustment coefficient, lag, spectrum residual shape vector, spectrum residual) Gain, scale factor).

  Further, a part of the encoded codes may be shared between adjacent bands without sending independent encoded codes in a plurality of bands. For example, when the bands Nn to Nw are divided into M bands and a common amplitude adjustment coefficient is used in two adjacent bands, the number of encoded codes of the amplitude adjustment coefficient is M / 2, and the other codes The number of conversion codes is M.

In the present embodiment, the case where a primary AR type pitch filter is used has been described. However, a filter to which the present invention can be applied is not limited to a first-order AR type pitch filter, and the present invention can also be applied to a filter whose transfer function is expressed by the following equation (5). The more the pitch filter having the parameters L and M that define the filter order is used, the more various characteristics can be expressed and the quality may be improved. However, since it becomes necessary to allocate more encoded bits of the filter coefficient as the order increases, it is necessary to determine an appropriate pitch filter transfer function based on practical bit allocation.

  In this embodiment, it is assumed that auditory masking is used, but a configuration that does not use auditory masking may be used. In this case, it is not necessary to provide the auditory masking calculation unit 204 of FIG. 2 in the second layer encoding unit 105, and the calculation amount of the entire apparatus can be reduced.

  Here, the configuration of the bit stream output from the multiplexing unit 104 will be described with reference to FIG. A first layer encoded code and a second layer encoded code are stored sequentially from the MSB (Most Significant Bit) of the bit stream. Further, the second layer encoded code is stored in the order of scale factor, amplitude adjustment coefficient, lag, spectral residual gain, spectral residual shape vector, and the latter information is arranged at a position closer to LSB (Least Significant Bit). ing. The configuration of this bit stream is high in code error sensitivity (deteriorates greatly) with respect to the sensitivity to code loss of each encoded code (how much the quality of the decoded signal is degraded when the encoded code is lost). The closer one is to the MSB. According to this configuration, when the bit stream is partially discarded on the transmission path, the degradation due to the discarding can be minimized by discarding sequentially from the LSB. In an example of a network configuration in which the bit stream is discarded preferentially from the LSB side, each encoded code divided as shown in FIG. 5 is transmitted in separate packets, prioritized for each packet, A configuration that uses a packet network that can be used. However, the network configuration is not limited to that described above.

  In addition, in the bit stream configuration in which the coding parameters with higher code error sensitivity are arranged closer to the MSB as shown in FIG. 5, channel coding is applied such that stronger error detection and error correction is applied to the bits closer to the MSB. By doing so, it is possible to obtain an effect that degradation of decoding quality can be minimized. For example, a CRC code, an RS code, or the like can be applied as a method for error detection and error correction.

  FIG. 6 is a block diagram illustrating a configuration of a decoding device 600 that forms, for example, a speech decoding device.

  Decoding apparatus 600 includes a separation unit 601 that separates the bitstream output from encoding apparatus 100 into a first layer encoded code and a second layer encoded code, and a first layer that decodes the first layer encoded code. It has the decoding part 602 and the 2nd layer decoding part 603 which decodes a 2nd layer encoding code.

  Separating section 601 receives the bitstream sent from encoding apparatus 100, separates it into a first layer encoded code and a second layer encoded code, and first layer decoding section 602 and second layer The data are output to the decryption unit 603, respectively.

  First layer decoding section 602 generates a first layer decoded signal from the first layer encoded code, and outputs the first layer decoded signal to second layer decoding section 603. In addition, the generated first layer decoded signal is output as a decoded signal (first layer decoded signal) with guaranteed minimum quality, if necessary.

  Second layer decoding section 603 generates a high-quality decoded signal (herein referred to as a second layer decoded signal) using the first layer decoded signal and the second layer encoded code, and if necessary, This decoded signal is output.

  In this way, the minimum quality of the reproduced sound is ensured by the first layer decoded signal, and the quality of the reproduced sound can be enhanced by the second layer decoded signal. In addition, whether the output signal is the first layer decoded signal or the second layer decoded signal depends on whether the second layer encoded code can be obtained depending on the network environment (occurrence of packet loss, etc.) Depends on user settings.

  The configuration of second layer decoding section 603 will be described in detail using FIG. In FIG. 7, second layer decoding section 603 has extended band decoding section 701, frequency domain transform section 702, and time domain transform section 703.

  Frequency domain transform section 702 transforms the first layer decoded signal input from first layer decoding section 602 into frequency domain parameters (for example, MDCT coefficients), and the parameters are subjected to first layer decoding with a spectrum score of Nn. The spectrum is output to extended band decoding section 701 as a spectrum.

  The extension band decoding unit 701 uses various parameters (amplitude adjustment coefficient, lag, spectrum residual shape vector, spectrum) from the second layer encoded code (in this configuration, the same as the extension band encoded code) input from the separation unit 601. (Residual gain, scale factor). Also, a second decoded spectrum that is band-extended using the various decoded parameters and the first layer decoded spectrum output from the frequency domain transform section 702 and having a spectrum score of Nw is generated. . Then, the second decoded spectrum is output to time domain transform section 703.

  The time domain conversion unit 703 converts the second decoded spectrum into a time domain signal, and then performs appropriate processing such as windowing and superposition addition as necessary to avoid discontinuities between frames. The second layer decoded signal is output.

  Next, detailed description of the extended band decoding unit 701 will be given with reference to FIG. In FIG. 8, the extended band decoding unit 701 includes a separation unit 801, an amplitude adjustment unit 802, a filter state setting unit 803, a filtering unit 804, a spectral residual shape codebook 805, a spectral residual gain codebook 806, and a multiplier 807. A scale factor decoding unit 808, a scaling unit 809, and a spectrum synthesis unit 810.

  The separation unit 801 converts the extension band encoded code input from the separation unit 601 into an amplitude adjustment coefficient encoded code, a lag encoded code, a residual shape encoded code, a residual gain encoded code, a scale factor encoded code, To separate. Also, the amplitude adjustment coefficient encoded code is stored in the amplitude adjuster 802, the lag encoded code is stored in the filtering unit 804, the residual shape encoded code is stored in the spectral residual shape codebook 805, and the residual gain encoded code is stored in the spectral residual. The scale factor encoded code is output to the difference gain codebook 806 to the scale factor decoding unit 808, respectively.

  The amplitude adjustment unit 802 decodes the amplitude adjustment coefficient encoded code input from the separation unit 801, and uses the decoded amplitude adjustment coefficient to separately detect the amplitude of the first layer decoded spectrum input from the frequency domain transform unit 702. The first layer decoded spectrum after amplitude adjustment is output to the filter state setting section 803. The amplitude adjustment is performed by the method represented by the above-described equation (1). Here, S1 (k) represents the first layer decoded spectrum, and S1 '(k) represents the first layer decoded spectrum after amplitude adjustment.

  The filter state setting unit 803 sets the first layer decoded spectrum after amplitude adjustment to the filter state of the pitch filter of the transfer function represented by the above equation (2). Specifically, the amplitude-adjusted first layer decoded spectrum {S1 ′ (k); 0 ≦ k <Nn} is substituted into the generated spectrum buffer S (k) and output to the filtering unit 804. Here, T is a pitch filter lag. The generated spectrum buffer S (k) is an array variable defined in a range of k = 0 to Nw−1, and a spectrum of (Nw−Nn) points is generated by this filtering process.

  The filtering unit 804 decodes the lag T input from the separation unit 801, and performs a filtering process on the generated spectrum buffer S (k) input from the filter state setting unit 803 using the decoded lag T. . Specifically, the output spectrum {S (k); Nn ≦ k <Nw} is generated by the method shown in the above-described equation (3). Here, g (j) represents a spectral residual gain represented by a residual gain encoding code j, and C (i, k) represents a spectral residual shape vector represented by a residual shape encoding code i. G (j) · C (i, k) is input from the multiplier 807. The generated output spectrum {S (k); Nn ≦ k <Nw} of the filtering unit 804 is output to the scaling unit 809.

  The spectral residual shape codebook 805 decodes the residual shape encoded code input from the separation unit 801 and outputs a spectral residual shape vector C (i, k) corresponding to the decoding result to the multiplier 807.

  The spectral residual gain codebook 806 decodes the residual gain encoded code input from the separation unit 801 and outputs a spectral residual gain g (j) corresponding to the decoding result to the multiplier 807.

  The multiplier 807 has a spectral residual shape vector C (i, k) input from the spectral residual shape codebook 805, a spectral residual gain g (j) input from the spectral residual gain codebook 806, and Is output to the filtering unit 804.

  The scale factor decoding unit 808 decodes the scale factor encoded code input from the separation unit 801 and outputs the decoded scale factor to the scaling unit 809.

The scaling unit 809 multiplies the output spectrum {S (k); Nn ≦ k <Nw} given from the filtering unit 804 by the scale factor inputted from the scale factor decoding unit 808, and spectrally synthesizes the multiplication result. Output to the unit 810.

  Spectrum synthesis section 810 includes first layer decoded spectrum {S1 (k); 0 ≦ k <Nn} given from frequency domain transform section 702, and a high-frequency section of the generated spectrum buffer after scaling output from scaling section 809 A spectrum obtained by combining {S (k); Nn ≦ k <Nw} is output to the time domain transforming unit 703 as a second decoded spectrum.

(Embodiment 2)
FIG. 9 shows the configuration of second layer encoding section 105 according to Embodiment 2 of the present invention. In FIG. 9, blocks having the same names as those in FIG. 2 have the same functions, and therefore detailed description thereof is omitted here. The difference between FIG. 2 and FIG. 9 is that a first spectrum encoding unit 901 exists between the frequency domain transform unit 201 and the extended band encoding unit 202. First spectrum encoding section 901 improves the quality of the first layer decoded spectrum output from frequency domain transform section 201, and the encoded code (first spectrum encoded code) at that time is sent to multiplexing section 104. The first layer decoded spectrum (first decoded spectrum) with improved quality is output to the extended band coding unit 202. The extension band encoding unit 202 performs the above-described processing using the first decoded spectrum, and outputs an extension band encoded code as a result. That is, the second layer encoded code of the present embodiment is a combination of the extended band encoded code and the first spectrum encoded code. Therefore, in the present embodiment, multiplexing section 104 multiplexes the first layer encoded code, the extended band encoded code, and the first spectrum encoded code to generate a bit stream.

  Next, details of the first spectrum encoding unit 901 will be described with reference to FIG. The first spectrum encoding unit 901 includes a scaling coefficient encoding unit 1001, a scaling coefficient decoding unit 1002, a fine spectrum encoding unit 1003, a multiplexing unit 1004, a fine spectrum decoding unit 1005, a normalization unit 1006, and a subtractor 1007. And an adder 1008.

  The subtractor 1007 generates a residual spectrum by subtracting the first layer decoded spectrum from the original spectrum, and outputs the residual spectrum to the scaling coefficient encoding unit 1001 and the normalizing unit 1006. The scaling coefficient encoding unit 1001 calculates a scaling coefficient representing the spectral outline of the residual spectrum, encodes the scaling coefficient, and outputs the encoded code to the multiplexing unit 1004 and the scaling coefficient decoding unit 1002.

  Auditory masking may be used in encoding the scaling factor. For example, bit allocation necessary for encoding the scaling coefficient is determined using auditory masking, and encoding is performed based on the bit allocation information. At this time, if there is a band to which no bits are allocated, the scaling coefficient of that band is not encoded. Thereby, the encoding of a scaling factor can be made efficient.

  The scaling coefficient decoding unit 1002 decodes the scaling coefficient from the input scaling coefficient encoded code, and outputs the decoded scaling coefficient to the normalization unit 1006, the fine spectrum encoding unit 1003, and the fine spectrum decoding unit 1005. .

  The normalization unit 1006 normalizes the residual spectrum provided from the subtractor 1007 using the scaling coefficient provided from the scaling coefficient decoding unit 1002, and the normalized residual spectrum is subjected to the fine spectrum encoding unit 1003. Output to.

The fine spectrum encoding unit 1003 calculates the auditory importance of each band using the scaling coefficient input from the scaling coefficient decoding unit 1002, obtains the number of bits allocated to each band, and determines the condition of the number of bits. Originally, the normalized residual spectrum (fine spectrum) is encoded. Then, the fine spectrum encoded code obtained by this encoding is output to multiplexing section 1004 and fine spectrum decoding section 1005.

  In addition, when encoding the residual spectrum after normalization, it may be encoded using auditory masking so as to reduce auditory distortion. Moreover, you may make it use the information of a 1st layer decoding spectrum for calculation of auditory importance. In this case, the first layer decoded spectrum is input to the fine spectrum encoding unit 1003.

  The encoded codes output from the scaling coefficient encoding unit 1001 and the fine spectrum encoding unit 1003 are multiplexed by the multiplexing unit 1004 and output to the multiplexing unit 104 as the first spectrum encoded code.

  The fine spectrum decoding unit 1005 calculates the auditory importance of each band using the scaling coefficient input from the scaling coefficient decoding unit 1002, obtains the number of bits assigned to each band, and calculates the scaling coefficient and the fine spectrum. The residual spectrum of each band is decoded from the fine spectrum encoded code input from encoding section 1003, and the decoded residual spectrum is output to adder 1008. In addition, you may make it use the information of a 1st layer decoding spectrum for calculation of auditory importance. In this case, the first layer decoded spectrum is configured to be input to the fine spectrum decoding unit 1005.

  Adder 1008 adds the decoded residual spectrum and the first layer decoded spectrum to generate a first decoded spectrum, and outputs the generated first decoded spectrum to extension band coding section 202.

  As described above, according to the present embodiment, after improving the quality of the first layer decoded spectrum, the extended band encoding unit 202 uses the spectrum after the quality improvement, that is, the first spectrum, in the high band part (Nn ≦ By generating a spectrum of k <Nw), it is possible to improve the quality of the band-extended decoded signal.

  The configuration of second layer decoding section 603 of the present embodiment will be described in detail using FIG. In FIG. 11, since the block having the same name as FIG. 7 has the same function, its detailed description is omitted here. In FIG. 11, second layer decoding section 603 has demultiplexing section 1101, first spectrum decoding section 1102, extended band decoding section 701, frequency domain transform section 702 and time domain transform section 703.

  Separating section 1101 separates the second layer encoded code into a first spectrum encoded code and an extended band encoded code, and sends the first spectrum encoded code to first spectrum decoding section 1102 The encoded code is output to extended band decoding section 701, respectively.

  Frequency domain transform section 702 transforms the first layer decoded signal input from first layer decoding section 602 into a frequency domain parameter (for example, MDCT coefficient), and uses this parameter as the first layer decoded spectrum as the first spectrum. The data is output to the decoding unit 1102.

  First spectrum decoding section 1102 receives from frequency domain transform section 702 the quantized spectrum of the first layer encoding error obtained by decoding the first spectrum encoded code input from demultiplexing section 1101. Added to the first layer decoded spectrum. Then, the addition result is output to extended band decoding section 701 as the first decoded spectrum.

Here, the first spectrum decoding section 1102 will be described in detail with reference to FIG. The first spectrum decoding unit 1102 includes a separation unit 1201, a scaling coefficient decoding unit 1202,
A fine spectrum decoding unit 1203 and a spectrum decoding unit 1204 are provided.

  The separation unit 1201 separates an encoded code representing a scaling coefficient and an encoded code representing a fine spectrum (spectral fine structure) from the input first spectrum encoded code, and obtains the scaling coefficient encoded code. The fine spectrum encoded code is output to scaling coefficient decoding section 1202 and fine spectrum decoding section 1203, respectively.

  The scaling coefficient decoding unit 1202 decodes the scaling coefficient from the input scaling coefficient encoded code, and outputs the decoded scaling coefficient to the spectrum decoding unit 1204 and the fine spectrum decoding unit 1203.

  The fine spectrum decoding unit 1203 calculates the auditory importance of each band using the scaling coefficient input from the scaling coefficient decoding unit 1202, and obtains the number of bits assigned to the fine spectrum of each band. In addition, the fine spectrum of each band is decoded from the fine spectrum encoded code input from separation section 1201, and the decoded fine spectrum is output to spectrum decoding section 1204.

  In addition, you may make it use the information of a 1st layer decoding spectrum for calculation of auditory importance. In that case, the first layer decoded spectrum is configured to be input to the fine spectrum decoding unit 1203.

  Spectrum decoding section 1204, first layer decoded spectrum provided from frequency domain transform section 702, scaling coefficient input from scaling coefficient decoding section 1202, fine spectrum input from the fine spectrum decoding section 1203, The first decoded spectrum is decoded, and this decoded spectrum is output to extended band decoding section 701.

The extended band encoding unit 202 of the present embodiment does not have to be provided with the spectrum residual shape codebook 305 and the spectrum residual gain codebook 307. The configuration of extension band encoding section 202 in this case is shown in FIG. In addition, extended band decoding section 701 does not have to provide spectral residual shape codebook 805 and spectral residual gain codebook 806. The configuration of extended band decoding section 701 in this case is shown in FIG. Note that the output signals of the filtering units 1301 and 1401 shown in FIGS. 13 and 14 are expressed by the following equation (6).

  In the present embodiment, after the quality of the first layer decoded spectrum is improved, a spectrum of a high frequency band (Nn ≦ k <Nw) is generated by extension band coding section 202 using the spectrum after the quality improvement. According to this configuration, the quality of the decoded signal can be improved. This advantage can be enjoyed regardless of the presence or absence of the spectral residual shape codebook and the spectral residual gain codebook.

In the first spectrum encoding unit 901, when encoding the spectrum of the low band part (0 ≦ k <Nn), the encoding distortion of the entire band (0 ≦ k <Nw) is minimized. You may encode the spectrum of a low-pass part (0 <= k <Nn). In this case, the extension band encoding unit 202 performs the encoding up to the high band part (Nn ≦ k <Nw). In this case, in the first spectrum encoding unit 901, the low band part is encoded in consideration of the influence of the low band part encoding result on the high band part encoding. Accordingly, since the spectrum of the low frequency band is encoded so that the spectrum of the entire band is optimized, the effect of improving the quality can be obtained.

(Embodiment 3)
The configuration of second layer coding section 105 according to Embodiment 3 of the present invention is shown in FIG. In FIG. 15, blocks having the same names as those in FIG. 9 have the same functions, and thus detailed description thereof is omitted here.

  The difference from FIG. 9 is that an extension band encoding unit 1501 that has a decoding function and obtains an extension band encoded code, generates a second decoded spectrum using the extension band encoded code, and generates a second decoded spectrum from the original spectrum. And a second spectrum encoding unit 1502 for encoding an error spectrum obtained by subtracting the decoded spectrum. By encoding the error spectrum described above with the second spectrum encoding unit 1502, a higher quality decoded spectrum can be generated, and the quality of the decoded signal obtained by the decoding apparatus can be improved.

  The extension band encoding unit 1501 generates and outputs an extension band encoded code in the same manner as the extension band encoding unit 202 shown in FIG. Also, the extended band encoding unit 1501 includes the same configuration as the extended band decoding unit 701 shown in FIG. 8, and generates the second decoded spectrum in the same manner as the extended band decoding unit 701. This second decoded spectrum is output to second spectrum encoding section 1502. That is, the second layer encoded code of the present embodiment includes an extended band encoded code, a first spectrum encoded code, and a second spectrum encoded code.

  In the configuration of extended band encoding section 1501, blocks having common names in FIGS. 3 and 8 may be shared.

  As shown in FIG. 16, the second spectrum encoding unit 1502 includes a scaling coefficient encoding unit 1601, a scaling coefficient decoding unit 1602, a fine spectrum encoding unit 1603, a multiplexing unit 1604, a normalization unit 1605, and a subtracter 1606. Have

  Subtractor 1606 generates a residual spectrum by subtracting the second decoded spectrum from the original spectrum, and outputs the residual spectrum to scaling coefficient encoding section 1601 and normalization section 1605. Scaling coefficient encoding section 1601 calculates a scaling coefficient that represents the spectral outline of the residual spectrum, encodes the scaling coefficient, and outputs the scaling coefficient encoded code to multiplexing section 1604 and scaling coefficient decoding section 1602. .

  Here, the efficiency of encoding the scaling coefficient may be improved by using auditory masking. For example, bit allocation necessary for encoding the scaling coefficient is determined using auditory masking, and encoding is performed based on the bit allocation information. At this time, if there is a band to which no bits are allocated, the scaling coefficient of that band is not encoded.

  The scaling coefficient decoding unit 1602 decodes the scaling coefficient from the input scaling coefficient encoded code, and outputs the decoded scaling coefficient to the normalization unit 1605 and the fine spectrum encoding unit 1603.

  The normalization unit 1605 normalizes the residual spectrum provided from the subtractor 1606 using the scaling coefficient provided from the scaling coefficient decoding unit 1602, and converts the residual spectrum after normalization into a fine spectrum encoding unit 1603. Output to.

The fine spectrum encoding unit 1603 calculates the aural importance of each band using the decoding scaling coefficient input from the scaling coefficient decoding unit 1602, obtains the number of bits allocated to each band, and calculates the number of bits. The residual spectrum (normal spectrum) after normalization is encoded under the conditions. Then, the encoded code obtained by this encoding is output to multiplexing section 1604.

  In addition, when encoding the residual spectrum after normalization, it may be encoded using auditory masking so as to reduce auditory distortion. Moreover, you may make it use the information of a 2nd layer decoding spectrum for calculation of auditory importance. In this case, the second layer decoded spectrum is configured to be input to the fine spectrum encoding unit 1603.

  The encoded codes output from the scaling coefficient encoding unit 1601 and the fine spectrum encoding unit 1603 are multiplexed by the multiplexing unit 1604 and output as the second spectrum encoded code.

  FIG. 17 illustrates a modification of the configuration of the second spectrum encoding unit 1502. In FIG. 17, since the block having the same name as that of FIG. 16 has the same function, detailed description thereof is omitted here.

  In this configuration, the second spectrum encoding unit 1502 directly encodes the residual spectrum given from the subtracter 1606. That is, normalization of the residual spectrum is not performed. Therefore, in this configuration, the scaling coefficient encoding unit 1601, the scaling coefficient decoding unit 1602, and the normalization unit 1605 shown in FIG. 16 are not provided. According to this configuration, it is not necessary for the second spectrum encoding unit 1502 to allocate bits to the scaling coefficient, so that the bit rate can be reduced.

  Auditory importance and bit allocation calculation section 1701 calculates the auditory importance of each band from the second decoded spectrum, and calculates the bit allocation to each band determined according to the auditory importance. The obtained auditory importance and bit allocation are output to the fine spectrum encoding unit 1603.

  The fine spectrum encoding unit 1603 encodes the residual spectrum based on the auditory importance and bit allocation input from the auditory importance and bit allocation calculating unit 1701. Then, the encoded code obtained by this encoding is output to the multiplexing unit 104 as the second spectrum encoded code. When encoding the residual spectrum, it may be encoded using auditory masking to reduce auditory distortion.

  The configuration of second layer decoding section 603 in the present embodiment is shown in FIG. Second layer decoding section 603 includes extension band decoding section 701, frequency domain transform section 702, time domain transform section 703, separation section 1101, first spectrum decoding section 1102 and second spectrum decoding section 1801. In FIG. 18, since the block having the same name as that of FIG. 11 has the same function, the detailed description thereof is omitted here.

  The second spectrum decoding unit 1801 quantizes the spectrum obtained by quantizing the coding error of the second decoded spectrum obtained by decoding the second spectrum encoded code input from the demultiplexing unit 1101, and the extended band decoding unit It adds to the 2nd decoding spectrum input from 701. FIG. Then, the addition result is output to the time domain conversion unit 703 as a third decoded spectrum.

  Second spectrum decoding section 1801 adopts the same configuration as that in FIG. 12 when second spectrum encoding section 1502 adopts the configuration shown in FIG. However, the first spectrum encoded code, the first layer decoded spectrum, and the first decoded spectrum in FIG. 12 are replaced with the second spectrum encoded code, the second decoded spectrum, and the third decoded spectrum, respectively.

  Further, in the present embodiment, the configuration of second spectrum decoding section 1801 has been described by taking the case where second spectrum encoding section 1502 adopts the configuration shown in FIG. 16 as an example, but second spectrum encoding section When 1502 adopts the configuration shown in FIG. 17, the configuration of second spectrum decoding section 1801 is as shown in FIG.

  That is, FIG. 19 shows the configuration of second spectrum decoding section 1801 corresponding to second spectrum encoding section 1502 that does not use a scaling coefficient. Second spectrum decoding section 1801 has auditory importance and bit allocation calculation section 1901, fine spectrum decoding section 1902, and spectrum decoding section 1903.

  In FIG. 19, the auditory importance and bit allocation calculating unit 1901 obtains the auditory importance of each band from the second decoded spectrum input from the extended band decoding unit 701, and is determined according to the auditory importance. Obtain the bit allocation to the band. The obtained auditory importance and bit allocation are output to the fine spectrum decoding unit 1902.

  The fine spectrum decoding unit 1902 performs fine spectrum encoding input from the separation unit 1101 as the second spectral encoding code based on the auditory importance and bit allocation input from the auditory importance and bit allocation calculation unit 1901. The code is decoded, and the decoding result (fine spectrum of each band) is output to the spectrum decoding unit 1903.

  The fine spectrum decoding unit 1903 adds the fine spectrum input from the fine spectrum decoding unit 1902 to the second decoded spectrum input from the extended band decoding unit 701, and adds the addition result to the third decoded spectrum. Output to the outside.

  In the present embodiment, the configuration including first spectrum encoding section 901 and first spectrum decoding section 1102 has been described as an example. However, first spectrum encoding section 901 and first spectrum decoding section 1102 are described. Even if there is no, the effect of this Embodiment is realizable. In this case, the configuration of second layer encoding section 105 is shown in FIG. 20, and the configuration of second layer decoding section 603 is shown in FIG.

  The embodiments of the scalable decoding device and the scalable encoding device according to the present invention have been described above.

  In the above-described embodiment, MDCT is used as the conversion method. However, the present invention is not limited to this, and the present invention is also applicable when other conversion methods such as Fourier transform, cosine transform, and Wavelet transform are used. it can.

  In the above-described embodiment, the description has been made based on the number of layers 2. However, the present invention is not limited to this.

  The encoding device and decoding device according to the present invention are not limited to the above-described first to third embodiments, and can be implemented with various modifications. For example, each embodiment can be implemented in combination as appropriate.

  The encoding device and the decoding device according to the present invention can also be mounted on a communication terminal device and a base station device in a mobile communication system, whereby the communication terminal device and the base station having the same operational effects as described above An apparatus can be provided.

Further, here, a case has been described as an example where the present invention is configured with hardware, but the present invention can also be implemented with software.

  Each functional block used in the description of each of the above embodiments is typically realized as an LSI that is an integrated circuit. These may be individually made into one chip, or may be made into one chip so as to include a part or all of them.

  The name used here is LSI, but it may also be called IC, system LSI, super LSI, or ultra LSI depending on the degree of integration.

  Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. An FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the LSI or a reconfigurable processor that can reconfigure the connection and setting of the circuit cells inside the LSI may be used.

  Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology. Biotechnology can be applied.

  That is, the scalable encoding device according to the above embodiment is a scalable encoding device that generates low frequency band encoded information and high frequency band encoded information from an original signal, and includes the low frequency band code. First spectrum calculating means for calculating a first spectrum in a low frequency band from the decoded signal of the digitized information, second spectrum calculating means for calculating a second spectrum from the original signal, the first spectrum and the second spectrum A first parameter calculating means for calculating a first parameter indicating the degree of similarity with the high frequency band, and a second parameter for calculating a second parameter indicating a fluctuation component between the high frequency band of the first spectrum and the second spectrum. Calculating means, and encoding means for encoding the calculated first parameter and second parameter as encoding information of the high frequency band, A configuration that.

  In the scalable encoding device according to the above-described embodiment, in the above configuration, the first parameter calculation unit uses a filter having the first spectrum as an internal state to set a parameter indicating a characteristic of the filter in the first configuration. A configuration for outputting as one parameter is adopted.

  In the scalable encoding device according to the above-described embodiment, the second parameter calculation unit includes a spectrum residual shape codebook in which a plurality of spectrum residual candidates are recorded, and the spectrum residual code has the configuration described above. A configuration is adopted in which the sign of the difference is output as the second parameter.

  The scalable coding apparatus according to the above embodiment further includes residual component encoding means for encoding the residual component between the first spectrum and the low frequency band portion of the second spectrum in the above configuration. The first parameter calculating means and the second parameter calculating means improve the quality of the first spectrum using the residual component encoded by the residual component encoding means, and then A configuration for calculating the parameter and the second parameter is adopted.

  In the scalable encoding device according to the above-described embodiment, in the above configuration, the residual component encoding unit includes the quality of the low frequency band portion of the first spectrum and the encoding unit encoded by the encoding unit. A configuration is adopted in which both the quality of the high frequency band portion of the decoded spectrum obtained from the first parameter and the second parameter are improved.

Further, the scalable coding apparatus according to the above embodiment has the above configuration in the first configuration.
The parameter includes a lag, the second parameter includes a spectrum residual, and further includes configuration means for configuring a bitstream arranged in the order of the lag and the spectrum residual.

  The scalable encoding device according to the above embodiment is an encoding device that generates low frequency band encoding information and high frequency band encoding information from an original signal, and the low frequency band encoding A first spectrum calculating means for calculating a first spectrum in a low frequency band from a decoded signal of information; a second spectrum calculating means for calculating a second spectrum from the original signal; and a high frequency of the first spectrum and the second spectrum. Parameter calculating means for calculating a parameter indicating the degree of similarity with the band part, parameter encoding means for encoding the calculated parameter as the encoding information of the high frequency band, low values of the first spectrum and the second spectrum A residual component encoding unit that encodes a residual component with the frequency band unit, and the parameter calculation unit includes the residual component encoding unit. After improving the quality of the first spectrum using the coded residual components Te, a configuration to calculate the parameters.

  Further, the scalable decoding device according to the embodiment includes a spectrum acquisition unit that acquires a first spectrum corresponding to a low frequency band, and a first parameter that is encoded as encoded information of a high frequency band, A first parameter indicating the degree of similarity between the first spectrum and the high-frequency band portion of the second spectrum corresponding to the original signal; and a second parameter encoded as high-frequency band encoding information, Parameter acquisition means for acquiring a second parameter indicating a fluctuation component with respect to the high-frequency band section; and decoding means for decoding the second spectrum using the acquired first parameter and second parameter. Take the configuration.

The scalable encoding method according to the above embodiment is a scalable encoding method for generating low frequency band encoded information and high frequency band encoded information from an original signal, and the low frequency band code A first spectrum calculating step for calculating a first spectrum in a low frequency band from the decoded signal of the conversion information, a second spectrum calculating step for calculating a second spectrum from the original signal, and the first spectrum and the second spectrum A first parameter calculating step for calculating a first parameter indicating the degree of similarity with the high frequency band part; and a second parameter for calculating a second parameter indicating a fluctuation component between the high frequency band part of the first spectrum and the second spectrum. A code for encoding the calculation step and the calculated first parameter and second parameter as encoding information of the high frequency band And the step,
It was made to have.

  The scalable decoding method according to the embodiment includes a spectrum acquisition step of acquiring a first spectrum corresponding to a low frequency band, and a first parameter encoded as encoded information of a high frequency band, A first parameter indicating the degree of similarity between the first spectrum and the high-frequency band portion of the second spectrum corresponding to the original signal; and a second parameter encoded as high-frequency band encoding information, A parameter acquisition step for acquiring a second parameter indicating a fluctuation component with respect to the high-frequency band section; and a decoding step for decoding the second spectrum using the acquired first parameter and second parameter. I did it.

In particular, a first scalable coding apparatus according to the present invention uses a filter having a first spectrum as an internal state to estimate a high-frequency part of a second spectrum, and encodes and sends filter information. It has a spectrum residual shape codebook in which a plurality of spectral residual candidates are recorded, and applies a spectral residual as an input signal of the filter to perform filtering to estimate a high frequency part of the second spectrum. By using the difference, it becomes possible to encode the high-frequency component of the second spectrum that cannot be expressed by the deformation of the first spectrum, so that the estimation performance of the high-frequency part of the second spectrum is improved and high quality is achieved. Is made.

  The second scalable encoding device according to the present invention encodes an error component between the low-frequency part of the second spectrum and the first spectrum to improve the quality of the first spectrum, and then the first spectrum. Is used to estimate the high-frequency part of the second spectrum using a filter having an internal state, and after improving the quality of the first spectrum with respect to the low-frequency part of the second spectrum, the first spectrum after quality improvement is used. By estimating the high frequency part of the second spectrum, the estimation performance is improved and the quality is improved.

  The third scalable coding apparatus according to the present invention also includes an estimated spectrum generated by estimating a high frequency part of the second spectrum using a filter having the first spectrum as an internal state, and a high frequency part of the second spectrum. The error component between the low-frequency part of the second spectrum and the first spectrum is reduced so as to reduce both the error component between the low-frequency part of the second spectrum and the error component between the low-frequency part of the second spectrum and the first spectrum. When encoding an error component between the first spectrum and the low-frequency part of the second spectrum, both qualities of the estimated spectrum of the first spectrum and the high-frequency part of the second spectrum are improved at the same time. Since the first spectrum is encoded, high quality can be realized.

  In the first to third scalable encoding devices, when the bit stream to be transmitted to the decoding device is generated by the encoding device, the bit stream includes at least a scale factor, a dynamic range adjustment coefficient, a lag. , And the bit stream may be configured in this order. As a result, the bit stream configuration is arranged closer to the MSB (Most Significant Bit) of the bit stream as the parameter having a greater influence on the quality of the decoded signal, and therefore any bit from the LSB (Least Significant Bit) of the bit stream Even if a bit is deleted at a position, an effect that quality degradation hardly occurs is obtained.

  This specification is based on Japanese Patent Application No. 2004-322959 filed on November 5, 2004. All this content is included here.

  The encoding apparatus, decoding apparatus, encoding method, and decoding method according to the present invention can be applied to scalable encoding / decoding and the like.

FIG. 1 is a block diagram showing a configuration of an encoding apparatus according to Embodiment 1 of the present invention. The block diagram which shows the structure of the 2nd layer encoding part which concerns on Embodiment 1 of this invention. FIG. 3 is a block diagram showing a configuration of an extension band encoding unit according to Embodiment 1 of the present invention. Schematic diagram showing a generated spectrum buffer processed by the filtering unit of the extension band encoding unit according to Embodiment 1 of the present invention Schematic diagram showing the contents of the bitstream output from the multiplexing unit of the encoding apparatus according to Embodiment 1 of the present invention. The block diagram which shows the structure of the decoding apparatus which concerns on Embodiment 1 of this invention. The block diagram which shows the structure of the 2nd layer decoding part which concerns on Embodiment 1 of this invention. FIG. 2 is a block diagram showing a configuration of an extended band decoding unit according to Embodiment 1 of the present invention. The block diagram which shows the structure of the 2nd layer encoding part which concerns on Embodiment 2 of this invention. The block diagram which shows the structure of the 1st spectrum encoding part which concerns on Embodiment 2 of this invention. The block diagram which shows the structure of the 2nd layer decoding part which concerns on Embodiment 2 of this invention. The block diagram which shows the structure of the 1st spectrum decoding part which concerns on Embodiment 2 of this invention. FIG. 9 is a block diagram showing a configuration of an extension band encoding unit according to Embodiment 2 of the present invention. FIG. 9 is a block diagram showing a configuration of an extended band decoding unit according to Embodiment 2 of the present invention. The block diagram which shows the structure of the 2nd layer encoding part which concerns on Embodiment 3 of this invention. The block diagram which shows the structure of the 2nd spectrum encoding part which concerns on Embodiment 3 of this invention. The block diagram which shows the modification of a structure of the 2nd spectrum encoding part which concerns on Embodiment 3 of this invention. The block diagram which shows the structure of the 2nd layer decoding part which concerns on Embodiment 3 of this invention. The block diagram which shows the modification of a structure of the 2nd spectrum decoding part which concerns on Embodiment 3 of this invention. Block diagram showing a modification of the configuration of the second layer encoding section according to Embodiment 3 of the present invention Block diagram showing a modification of the configuration of the second layer decoding section according to Embodiment 3 of the present invention

Claims (5)

An encoding device for generating low frequency band encoded information and high frequency band encoded information from an original signal,
First spectrum calculating means for calculating a first spectrum of a low frequency band from a decoded signal of the encoded information of the low frequency band;
Second spectrum calculating means for calculating a second spectrum from the original signal;
Using a filter having the first spectrum as an internal state, first outputs a parameter indicating a characteristic of the filter, as a first parameter indicating a similar degree of high-frequency band portion of the second spectrum and the first spectrum Parameter calculation means;
One spectral residual candidate code from a spectral residual shape codebook in which a plurality of spectral residual candidates are recorded indicates a fluctuation component between the high frequency band portion of the first spectrum and the second spectrum. Second parameter calculating means for outputting as a second parameter;
Determining means for simultaneously determining the first parameter and the second parameter for generating an estimated value most similar to the high-frequency band portion of the second spectrum from the output first parameter and second parameter;
Encoding means for encoding the determined first parameter and second parameter as encoding information of the high frequency band;
An encoding device. A residual component encoding means for encoding a residual component between the first spectrum and the low frequency band of the second spectrum;
The first parameter calculation means and the second parameter calculation means are:
Calculating the first parameter and the second parameter after improving the quality of the first spectrum using the residual component encoded by the residual component encoding means;
The encoding device according to claim 1. The residual component encoding means includes:
Improving both the quality of the low frequency band portion of the first spectrum and the quality of the high frequency band portion of the decoded spectrum obtained from the first parameter and the second parameter encoded by the encoding means;
The encoding device according to claim 2 . The first parameter includes a lag, the second parameter includes a spectral residual;
And further comprising means for configuring a bitstream arranged in the order of the lag and the spectral residuals.
The encoding device according to claim 1. An encoding method for generating low frequency band encoded information and high frequency band encoded information from an original signal,
A first spectrum calculating step of calculating a first spectrum of a low frequency band from a decoded signal of the encoded information of the low frequency band;
A second spectrum calculating step of calculating a second spectrum from the original signal;
Using a filter having the first spectrum as an internal state, first to calculate the parameter indicating the characteristic of the filter, as a first parameter indicating a similar degree of high-frequency band portion of the second spectrum and the first spectrum A parameter calculation step;
One spectral residual candidate code from a spectral residual shape codebook in which a plurality of spectral residual candidates are recorded indicates a fluctuation component between the high frequency band portion of the first spectrum and the second spectrum. A second parameter calculating step for calculating as a second parameter;
A determination step of simultaneously determining the first parameter and the second parameter that generate an estimated value that is most similar to the high frequency band portion of the second spectrum from the calculated first parameter and second parameter;
An encoding step of encoding the determined first parameter and second parameter as encoding information of the high frequency band;
An encoding method comprising:

Images